UltraLogic: Enhancing LLM Reasoning through Large-Scale Data Synthesis and Bipolar Float Reward
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2026
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866914236757704704 |
|---|---|
| author | Liu, Yile Liu, Yixian Li, Zongwei Huang, Yufei Feng, Xinhua Hu, Zhichao Hu, Jinglu Yan, Jianfeng Lian, Fengzong Liu, Yuhong |
| author_facet | Liu, Yile Liu, Yixian Li, Zongwei Huang, Yufei Feng, Xinhua Hu, Zhichao Hu, Jinglu Yan, Jianfeng Lian, Fengzong Liu, Yuhong |
| contents | While Large Language Models (LLMs) have demonstrated significant potential in natural language processing , complex general-purpose reasoning requiring multi-step logic, planning, and verification remains a critical bottleneck. Although Reinforcement Learning with Verifiable Rewards (RLVR) has succeeded in specific domains , the field lacks large-scale, high-quality, and difficulty-calibrated data for general reasoning. To address this, we propose UltraLogic, a framework that decouples the logical core of a problem from its natural language expression through a Code-based Solving methodology to automate high-quality data production. The framework comprises hundreds of unique task types and an automated calibration pipeline across ten difficulty levels. Furthermore, to mitigate binary reward sparsity and the Non-negative Reward Trap, we introduce the Bipolar Float Reward (BFR) mechanism, utilizing graded penalties to effectively distinguish perfect responses from those with logical flaws. Our experiments demonstrate that task diversity is the primary driver for reasoning enhancement , and that BFR, combined with a difficulty matching strategy, significantly improves training efficiency, guiding models toward global logical optima. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_03205 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | UltraLogic: Enhancing LLM Reasoning through Large-Scale Data Synthesis and Bipolar Float Reward Liu, Yile Liu, Yixian Li, Zongwei Huang, Yufei Feng, Xinhua Hu, Zhichao Hu, Jinglu Yan, Jianfeng Lian, Fengzong Liu, Yuhong Computation and Language Artificial Intelligence While Large Language Models (LLMs) have demonstrated significant potential in natural language processing , complex general-purpose reasoning requiring multi-step logic, planning, and verification remains a critical bottleneck. Although Reinforcement Learning with Verifiable Rewards (RLVR) has succeeded in specific domains , the field lacks large-scale, high-quality, and difficulty-calibrated data for general reasoning. To address this, we propose UltraLogic, a framework that decouples the logical core of a problem from its natural language expression through a Code-based Solving methodology to automate high-quality data production. The framework comprises hundreds of unique task types and an automated calibration pipeline across ten difficulty levels. Furthermore, to mitigate binary reward sparsity and the Non-negative Reward Trap, we introduce the Bipolar Float Reward (BFR) mechanism, utilizing graded penalties to effectively distinguish perfect responses from those with logical flaws. Our experiments demonstrate that task diversity is the primary driver for reasoning enhancement , and that BFR, combined with a difficulty matching strategy, significantly improves training efficiency, guiding models toward global logical optima. |
| title | UltraLogic: Enhancing LLM Reasoning through Large-Scale Data Synthesis and Bipolar Float Reward |
| topic | Computation and Language Artificial Intelligence |
| url | https://arxiv.org/abs/2601.03205 |